def initialize(self, comm, config, args):
rank = comm.Get_rank()
self.config = config
self.args = args
struct_file = reader.open(args.struct_file)
self.struct_filename = struct_file.filename
self.npoints = struct_file.nlines
self.idxs_thread = p_index.get_idxs_thread(comm, self.npoints)
if hasattr(struct_file, '_skip'): # multi-thread reading
coords_thread = struct_file.readlines(self.idxs_thread)
self.coords = np.vstack(comm.allgather(coords_thread))
else: # serial reading
if rank == 0:
self.coords = struct_file.readlines()
else:
self.coords = None
self.coords = comm.bcast(self.coords, root=0)
logging.info('input coordinates loaded')
self.initialize_local_scale()
self.initialize_weights()
self.initialize_metric()
self.neigs = 10
def get_weights(self, comm):
rank = comm.Get_rank()
size = comm.Get_size()
# compute the distance matrix if needed
if self.distance_matrix is None:
logging.info("distance matrix not provided, computing it...")
#idxs_thread = p_index.get_idxs_thread(comm, self.npoints)
#npoints_thread = len(idxs_thread)
#coords_thread = np.array([self.coords[idx] for idx in idxs_thread])
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(comm, self.npoints)
npoints_thread = self.npoints_per_thread[rank]
coords_thread = np.array([self.coords[idx] for idx in self.idxs_thread])
DistanceMatrix = mt.DistanceMatrix(coords_thread, self.coords, metric=self.metric, metric_prms=self.metric_prms)
self.distance_matrix = np.vstack(comm.allgather(DistanceMatrix.distance_matrix))
logging.info("distance matrix computed")
else:
if any(self.distance_matrix.shape[idx] != self.npoints for idx in [0, 1]):
logging.error("distance matrix provided doesn't match the number of coordinates")
self.sigma = self.initialize_sigma_values(self.sigma, self.ksigma)
# invert kernel matrix
kernel_matrix = self.fit_function(self.distance_matrix, self.sigma)
# perform singular valu decomposition
U, s, V = np.linalg.svd(kernel_matrix)
sinv = 1/s
# filter noisy singular values values
sinv[s<0.02*np.max(s)] = 0
inverse_kernel_matrix = np.dot(np.dot(V.T, np.diag(sinv)), U.T)
weights = np.dot(inverse_kernel_matrix, self.values)
logging.info("kernel matrix inverted")
return weights
def initialize(self, comm, config, args):
rank = comm.Get_rank()
size = comm.Get_size()
self.config = config
self.args = args
filename = args.struct_file[0]
self.struct_filename = filename
if rank == 0:
f = x2h.Fasu(args.topfile, args.struct_file, selection=args.selection)
c = x2h.Cofasu(f)
self.coords = np.array(c.x, dtype=np.double) * 0.1
self.coords = np.swapaxes(self.coords, 1, 2)
else:
self.coords = None
self.coords = comm.bcast(self.coords, root=0)
print rank, self.coords.shape
self.npoints = self.coords.shape[0]
self.natoms = self.coords.shape[2]
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(comm, self.npoints)
logging.info('input coordinates loaded')
self.initialize_local_scale()
self.initialize_weights()
self.initialize_metric()
def initialize(self, comm, config, args):
rank = comm.Get_rank()
size = comm.Get_size()
self.config = config
self.args = args
filename = args.struct_file[0]
self.struct_filename = filename
self.npoints,self.natoms = coord_reader.get_nframes_natoms(filename)
if coord_reader.supports_parallel_reading(filename):
# read coordinates in parallel
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(comm, self.npoints)
coords_thread = coord_reader.get_coordinates(filename, idxs=self.idxs_thread)
coords_ravel = coords_thread.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(self.npoints_per_thread,self.natoms)
coordstemp = np.zeros(self.npoints*3*self.natoms, dtype='float')
start = MPI.Wtime()
comm.Allgatherv(coords_ravel, (coordstemp, ravel_lengths, ravel_offsets, MPI.DOUBLE))
self.coords = coordstemp.reshape((self.npoints,3,self.natoms))
else:
# serial reading
if rank == 0:
self.coords = coord_reader.get_coordinates(filename)
else:
self.coords = np.zeros((self.npoints,3,self.natoms),dtype=np.double)
comm.Bcast(self.coords, root=0)
logging.info('input coordinates loaded')
self.initialize_local_scale()
self.initialize_weights()
self.initialize_metric()
def get_weights(self, comm):
rank = comm.Get_rank()
size = comm.Get_size()
# compute the distance matrix if needed
if self.distance_matrix is None:
logging.info("distance matrix not provided, computing it...")
#idxs_thread = p_index.get_idxs_thread(comm, self.npoints)
#npoints_thread = len(idxs_thread)
#coords_thread = np.array([self.coords[idx] for idx in idxs_thread])
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(
comm, self.npoints)
npoints_thread = self.npoints_per_thread[rank]
coords_thread = np.array(
[self.coords[idx] for idx in self.idxs_thread])
DistanceMatrix = mt.DistanceMatrix(coords_thread,
self.coords,
metric=self.metric,
metric_prms=self.metric_prms)
self.distance_matrix = np.vstack(
comm.allgather(DistanceMatrix.distance_matrix))
logging.info("distance matrix computed")
else:
if any(self.distance_matrix.shape[idx] != self.npoints
for idx in [0, 1]):
logging.error(
"distance matrix provided doesn't match the number of coordinates"
)
self.sigma = self.initialize_sigma_values(self.sigma, self.ksigma)
# invert kernel matrix
kernel_matrix = self.fit_function(self.distance_matrix, self.sigma)
# perform singular valu decomposition
U, s, V = np.linalg.svd(kernel_matrix)
sinv = 1 / s
# filter noisy singular values values
sinv[s < 0.02 * np.max(s)] = 0
inverse_kernel_matrix = np.dot(np.dot(V.T, np.diag(sinv)), U.T)
weights = np.dot(inverse_kernel_matrix, self.values)
logging.info("kernel matrix inverted")
return weights
def initialize(self, comm, config, args):
rank = comm.Get_rank()
size = comm.Get_size()
self.config = config
self.args = args
filename = args.struct_file[0]
self.struct_filename = filename
self.npoints, self.natoms = coord_reader.get_nframes_natoms(filename)
if coord_reader.supports_parallel_reading(filename):
# read coordinates in parallel
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(
comm, self.npoints)
coords_thread = coord_reader.get_coordinates(filename,
idxs=self.idxs_thread)
coords_ravel = coords_thread.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(
self.npoints_per_thread, self.natoms)
coordstemp = np.zeros(self.npoints * 3 * self.natoms,
dtype='float')
start = MPI.Wtime()
comm.Allgatherv(
coords_ravel,
(coordstemp, ravel_lengths, ravel_offsets, MPI.DOUBLE))
self.coords = coordstemp.reshape((self.npoints, 3, self.natoms))
else:
# serial reading
if rank == 0:
self.coords = coord_reader.get_coordinates(filename)
else:
self.coords = np.zeros((self.npoints, 3, self.natoms),
dtype=np.double)
comm.Bcast(self.coords, root=0)
logging.info('input coordinates loaded')
self.initialize_local_scale()
self.initialize_weights()
self.initialize_metric()
def initialize(self, comm, config, args):
rank = comm.Get_rank()
size = comm.Get_size()
self.config = config
self.args = args
filename = args.struct_file[0]
self.struct_filename = filename
self.npoints,self.natoms = coord_reader.get_nframes_natoms(filename)
if coord_reader.supports_parallel_reading(filename):
# read coordinates in parallel
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(comm, self.npoints)
coords_thread = coord_reader.get_coordinates(filename, idxs=self.idxs_thread)
coords_ravel = coords_thread.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(self.npoints_per_thread,self.natoms)
coordstemp = np.zeros(self.npoints*3*self.natoms, dtype='float')
start = MPI.Wtime()
comm.Allgatherv(coords_ravel, (coordstemp, ravel_lengths, ravel_offsets, MPI.DOUBLE))
self.coords = coordstemp.reshape((self.npoints,3,self.natoms))
else:
# serial reading
if rank == 0:
self.coords = coord_reader.get_coordinates(filename)
else:
self.coords = np.zeros((self.npoints,3,self.natoms),dtype=np.double)
comm.Bcast(self.coords, root=0)
logging.info('input coordinates loaded')
# load file of values
valfile = reader.open(args.valfile)
self.values = valfile.readlines()
format = os.path.splitext(args.valfile)[1]
if format == '.ev':
self.fitdcs = True
else:
self.fitdcs = False
if len(self.values.shape) > 2:
raise ValueError('file of values should contain a single column')
self.initialize_metric()
self.function = config.get(self.args.section,'function')
self.status_sigma = config.get(self.args.section,'status')
if self.status_sigma == 'constant':
self.sigma = config.getfloat(self.args.section,'sigma')
self.ksigma = None
elif self.status_sigma == 'kneighbor':
self.sigma = None
self.ksigma = config.getint(self.args.section,'ksigma')
if args.embed_file is not None:
#embed_file = reader.open(args.embed_file)
#self.embed_filename = embed_file.filename
#self.npoints_embed = embed_file.nlines
#self.idxs_thread_embed = p_index.get_idxs_thread_v(comm, self.npoints_embed)
#if hasattr(embed_file, '_skip'): # multi-thread reading
# coords_thread_embed = embed_file.readlines(self.idxs_thread_embed)
# self.coords_embed = np.vstack(comm.allgather(coords_thread_embed))
#else: # serial reading
# if rank == 0:
# self.coords_embed = embed_file.readlines()
# else:
# self.coords_embed = None
# self.coords_embed = comm.bcast(self.coords_embed, root=0)
filename_embed = args.embed_file
self.embed_filename = filename_embed
self.npoints_embed,self.natoms_embed = coord_reader.get_nframes_natoms(filename_embed)
if coord_reader.supports_parallel_reading(filename_embed):
# read coordinates in parallel
self.idxs_thread_embed, self.npoints_per_thread_embed, self.offsets_per_thread_embed = p_index.get_idxs_thread(comm, self.npoints_embed)
coords_thread_embed = coord_reader.get_coordinates(filename_embed, idxs=self.idxs_thread_embed)
coords_ravel_embed = coords_thread_embed.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(self.npoints_per_thread_embed,self.natoms_embed)
coordstemp_embed = np.zeros(self.npoints_embed*3*self.natoms_embed, dtype='float')
comm.Allgatherv(coords_ravel, (coordstemp_embed, ravel_lengths, ravel_offsets, MPI.DOUBLE))
self.coords_embed = coordstemp_embed.reshape((self.npoints_embed,3,self.natoms_embed))
else:
# serial reading
if rank == 0:
self.coords_embed = coord_reader.get_coordinates(filename_embed)
else:
self.coords_embed = np.zeros((self.npoints_embed,3,self.natoms_embed),dtype=np.double)
comm.Bcast(self.coords_embed, root=0)
def initialize(self, comm, config, args):
rank = comm.Get_rank()
size = comm.Get_size()
self.config = config
self.args = args
filename = args.struct_file[0]
self.struct_filename = filename
self.npoints, self.natoms = coord_reader.get_nframes_natoms(filename)
if coord_reader.supports_parallel_reading(filename):
# read coordinates in parallel
self.idxs_thread, self.npoints_per_thread, self.offsets_per_thread = p_index.get_idxs_thread(
comm, self.npoints)
coords_thread = coord_reader.get_coordinates(filename,
idxs=self.idxs_thread)
coords_ravel = coords_thread.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(
self.npoints_per_thread, self.natoms)
coordstemp = np.zeros(self.npoints * 3 * self.natoms,
dtype='float')
start = MPI.Wtime()
comm.Allgatherv(
coords_ravel,
(coordstemp, ravel_lengths, ravel_offsets, MPI.DOUBLE))
self.coords = coordstemp.reshape((self.npoints, 3, self.natoms))
else:
# serial reading
if rank == 0:
self.coords = coord_reader.get_coordinates(filename)
else:
self.coords = np.zeros((self.npoints, 3, self.natoms),
dtype=np.double)
comm.Bcast(self.coords, root=0)
logging.info('input coordinates loaded')
# load file of values
valfile = reader.open(args.valfile)
self.values = valfile.readlines()
format = os.path.splitext(args.valfile)[1]
if format == '.ev':
self.fitdcs = True
else:
self.fitdcs = False
if len(self.values.shape) > 2:
raise ValueError(
'file of values should contain a single column')
self.initialize_metric()
self.function = config.get(self.args.section, 'function')
self.status_sigma = config.get(self.args.section, 'status')
if self.status_sigma == 'constant':
self.sigma = config.getfloat(self.args.section, 'sigma')
self.ksigma = None
elif self.status_sigma == 'kneighbor':
self.sigma = None
self.ksigma = config.getint(self.args.section, 'ksigma')
if args.embed_file is not None:
#embed_file = reader.open(args.embed_file)
#self.embed_filename = embed_file.filename
#self.npoints_embed = embed_file.nlines
#self.idxs_thread_embed = p_index.get_idxs_thread_v(comm, self.npoints_embed)
#if hasattr(embed_file, '_skip'): # multi-thread reading
# coords_thread_embed = embed_file.readlines(self.idxs_thread_embed)
# self.coords_embed = np.vstack(comm.allgather(coords_thread_embed))
#else: # serial reading
# if rank == 0:
# self.coords_embed = embed_file.readlines()
# else:
# self.coords_embed = None
# self.coords_embed = comm.bcast(self.coords_embed, root=0)
filename_embed = args.embed_file
self.embed_filename = filename_embed
self.npoints_embed, self.natoms_embed = coord_reader.get_nframes_natoms(
filename_embed)
if coord_reader.supports_parallel_reading(filename_embed):
# read coordinates in parallel
self.idxs_thread_embed, self.npoints_per_thread_embed, self.offsets_per_thread_embed = p_index.get_idxs_thread(
comm, self.npoints_embed)
coords_thread_embed = coord_reader.get_coordinates(
filename_embed, idxs=self.idxs_thread_embed)
coords_ravel_embed = coords_thread_embed.ravel()
ravel_lengths, ravel_offsets = p_index.get_ravel_offsets(
self.npoints_per_thread_embed, self.natoms_embed)
coordstemp_embed = np.zeros(self.npoints_embed * 3 *
self.natoms_embed,
dtype='float')
comm.Allgatherv(coords_ravel, (coordstemp_embed, ravel_lengths,
ravel_offsets, MPI.DOUBLE))
self.coords_embed = coordstemp_embed.reshape(
(self.npoints_embed, 3, self.natoms_embed))
else:
# serial reading
if rank == 0:
self.coords_embed = coord_reader.get_coordinates(
filename_embed)
else:
self.coords_embed = np.zeros(
(self.npoints_embed, 3, self.natoms_embed),
dtype=np.double)
comm.Bcast(self.coords_embed, root=0)
